/* The weight variable */
const struct variable *wv;
+
};
+/* Per category data */
+struct descriptive_data
+{
+ const struct variable *var;
+ struct moments1 *mom;
+
+ double minimum;
+ double maximum;
+};
/* Workspace variable for each dependent variable */
struct per_var_ws
int n_groups;
- double cc;
+ double mse;
};
struct oneway_workspace
struct per_var_ws *vws;
- struct moments1 *totals;
- double minimum;
- double maximum;
+ /* An array of descriptive data. One for each dependent variable */
+ struct descriptive_data **dd_total;
};
/* Routines to show the output tables */
struct casereader *group;
bool ok;
-
-
grouper = casegrouper_create_splits (proc_open (ds), dict);
while (casegrouper_get_next_group (grouper, &group))
run_oneway (&oneway, group, ds);
static void postcalc (const struct oneway_spec *cmd);
static void precalc (const struct oneway_spec *cmd);
-struct descriptive_data
-{
- struct moments1 *mom;
- double minimum;
- double maximum;
-};
-
-static void *
-makeit (void)
+static struct descriptive_data *
+dd_create (const struct variable *var)
{
struct descriptive_data *dd = xmalloc (sizeof *dd);
+
dd->mom = moments1_create (MOMENT_VARIANCE);
dd->minimum = DBL_MAX;
dd->maximum = -DBL_MAX;
+ dd->var = var;
+
+ return dd;
+}
+
+
+static void *
+makeit (void *aux1, void *aux2 UNUSED)
+{
+ const struct variable *var = aux1;
+
+ struct descriptive_data *dd = dd_create (var);
return dd;
}
static void
-updateit (void *user_data, const struct variable *wv,
- const struct variable *catvar, const struct ccase *c, void *aux)
+updateit (void *user_data,
+ enum mv_class exclude,
+ const struct variable *wv,
+ const struct variable *catvar UNUSED,
+ const struct ccase *c,
+ void *aux1, void *aux2)
{
- const union value *val = case_data_idx (c, 0);
struct descriptive_data *dd = user_data;
- struct oneway_workspace *ws = aux;
- double weight = 1.0;
- if (wv)
- weight = case_data (c, wv)->f;
+ const struct variable *varp = aux1;
+
+ const union value *valx = case_data (c, varp);
+
+ struct descriptive_data *dd_total = aux2;
- moments1_add (dd->mom, val->f, weight);
- moments1_add (ws->totals, val->f, weight);
+ double weight;
- if (val->f * weight < dd->minimum)
- dd->minimum = val->f * weight;
+ if ( var_is_value_missing (varp, valx, exclude))
+ return;
- if (val->f * weight > dd->maximum)
- dd->maximum = val->f * weight;
+ weight = wv != NULL ? case_data (c, wv)->f : 1.0;
+ moments1_add (dd->mom, valx->f, weight);
+ if (valx->f * weight < dd->minimum)
+ dd->minimum = valx->f * weight;
- if (val->f * weight < ws->minimum)
- ws->minimum = val->f * weight;
+ if (valx->f * weight > dd->maximum)
+ dd->maximum = valx->f * weight;
+
+ {
+ const struct variable *var = dd_total->var;
+ const union value *val = case_data (c, var);
- if (val->f * weight > ws->maximum)
- ws->maximum = val->f * weight;
+ moments1_add (dd_total->mom,
+ val->f,
+ weight);
+
+ if (val->f * weight < dd_total->minimum)
+ dd_total->minimum = val->f * weight;
+
+ if (val->f * weight > dd_total->maximum)
+ dd_total->maximum = val->f * weight;
+ }
}
static void
struct oneway_workspace ws;
+ ws.actual_number_of_groups = 0;
ws.vws = xmalloc (cmd->n_vars * sizeof (*ws.vws));
+ ws.dd_total = xmalloc (sizeof (struct descriptive_data) * cmd->n_vars);
- ws.totals = moments1_create (MOMENT_VARIANCE);
- ws.minimum = DBL_MAX;
- ws.maximum = -DBL_MAX;
-
+ for (v = 0 ; v < cmd->n_vars; ++v)
+ ws.dd_total[v] = dd_create (cmd->vars[v]);
for (v = 0; v < cmd->n_vars; ++v)
{
struct categoricals *cats = categoricals_create (&cmd->indep_var, 1,
cmd->wv, cmd->exclude,
makeit,
- updateit, &ws);
+ updateit,
+ cmd->vars[v], ws.dd_total[v]);
ws.vws[v].cov = covariance_2pass_create (1, &cmd->vars[v],
cats,
cmd->wv, cmd->exclude);
- ws.vws[v].cc = 0;
}
c = casereader_peek (input, 0);
for (i = 0; i < cmd->n_vars; ++i)
{
- {
- struct per_var_ws *pvw = &ws.vws[i];
-
- pvw->cc += weight;
- covariance_accumulate_pass1 (pvw->cov, c);
- }
-
+ struct per_var_ws *pvw = &ws.vws[i];
const struct variable *v = cmd->vars[i];
-
const union value *val = case_data (c, v);
-
struct group_proc *gp = group_proc_get (cmd->vars[i]);
struct hsh_table *group_hash = gp->group_hash;
-
struct group_statistics *gs;
+ if ( MISS_ANALYSIS == cmd->missing_type)
+ {
+ if ( var_is_value_missing (v, val, cmd->exclude))
+ continue;
+ }
+
+ covariance_accumulate_pass1 (pvw->cov, c);
+
gs = hsh_find (group_hash, indep_val );
if ( ! gs )
for (i = 0; i < cmd->n_vars; ++i)
{
struct per_var_ws *pvw = &ws.vws[i];
+ const struct variable *v = cmd->vars[i];
+ const union value *val = case_data (c, v);
+
+ if ( MISS_ANALYSIS == cmd->missing_type)
+ {
+ if ( var_is_value_missing (v, val, cmd->exclude))
+ continue;
+ }
+
covariance_accumulate_pass2 (pvw->cov, c);
}
}
gsl_matrix *cm = covariance_calculate_unnormalized (pvw->cov);
const struct categoricals *cats = covariance_get_categoricals (pvw->cov);
+ double n;
+ moments1_calculate (ws.dd_total[v]->mom, &n, NULL, NULL, NULL, NULL);
+
pvw->sst = gsl_matrix_get (cm, 0, 0);
+ // gsl_matrix_fprintf (stdout, cm, "%g ");
+
reg_sweep (cm, 0);
pvw->sse = gsl_matrix_get (cm, 0, 0);
pvw->ssa = pvw->sst - pvw->sse;
pvw->n_groups = categoricals_total (cats);
+
+ pvw->mse = (pvw->sst - pvw->ssa) / (n - pvw->n_groups);
}
postcalc (cmd);
+
for (v = 0; v < cmd->n_vars; ++v)
{
struct categoricals *cats = covariance_get_categoricals (ws.vws[v].cov);
categoricals_done (cats);
+
+ if (categoricals_total (cats) > ws.actual_number_of_groups)
+ ws.actual_number_of_groups = categoricals_total (cats);
}
if ( cmd->stats & STATS_HOMOGENEITY )
casereader_destroy (input);
- ws.actual_number_of_groups = hsh_count (ws.group_hash);
-
if (!taint_has_tainted_successor (taint))
output_oneway (cmd, &ws);
show_anova_table (cmd, ws);
-
if (ll_count (&cmd->contrast_list) > 0)
{
show_contrast_coeffs (cmd, ws);
show_contrast_tests (cmd, ws);
}
-
/* Clean up */
for (i = 0; i < cmd->n_vars; ++i )
{
for (i = 0; i < cmd->n_vars; ++i)
{
+ double n;
+ double df1, df2;
+ double msa;
+ const char *s = var_to_string (cmd->vars[i]);
const struct per_var_ws *pvw = &ws->vws[i];
- struct group_proc *gp = group_proc_get (cmd->vars[i]);
- const double df1 = pvw->n_groups - 1;
- const double df2 = pvw->cc - pvw->n_groups;
- const double msa = pvw->ssa / df1;
- const char *s = var_to_string (cmd->vars[i]);
+ moments1_calculate (ws->dd_total[i]->mom, &n, NULL, NULL, NULL, NULL);
+
+ df1 = pvw->n_groups - 1;
+ df2 = n - pvw->n_groups;
+ msa = pvw->ssa / df1;
tab_text (t, 0, i * 3 + 1, TAB_LEFT | TAT_TITLE, s);
tab_text (t, 1, i * 3 + 1, TAB_LEFT | TAT_TITLE, _("Between Groups"));
tab_hline (t, TAL_1, 0, n_cols - 1, i * 3 + 1);
- gp->mse = (pvw->sst - pvw->ssa) / df2;
-
/* Sums of Squares */
tab_double (t, 2, i * 3 + 1, 0, pvw->ssa, NULL);
tab_double (t, 2, i * 3 + 3, 0, pvw->sst, NULL);
/* Degrees of freedom */
tab_fixed (t, 3, i * 3 + 1, 0, df1, 4, 0);
tab_fixed (t, 3, i * 3 + 2, 0, df2, 4, 0);
- tab_fixed (t, 3, i * 3 + 3, 0, pvw->cc - 1, 4, 0);
+ tab_fixed (t, 3, i * 3 + 3, 0, n - 1, 4, 0);
/* Mean Squares */
tab_double (t, 4, i * 3 + 1, TAB_RIGHT, msa, NULL);
- tab_double (t, 4, i * 3 + 2, TAB_RIGHT, gp->mse, NULL);
+ tab_double (t, 4, i * 3 + 2, TAB_RIGHT, pvw->mse, NULL);
{
- const double F = msa / gp->mse ;
+ const double F = msa / pvw->mse ;
/* The F value */
tab_double (t, 5, i * 3 + 1, 0, F, NULL);
{
double T;
double n, mean, variance;
+ double std_dev, std_error ;
+
+ struct string vstr;
const union value *gval = categoricals_get_value_by_subscript (cats, count);
const struct descriptive_data *dd = categoricals_get_user_data_by_subscript (cats, count);
moments1_calculate (dd->mom, &n, &mean, &variance, NULL, NULL);
- double std_dev = sqrt (variance);
- double std_error = std_dev / sqrt (n) ;
-
- struct string vstr;
+ std_dev = sqrt (variance);
+ std_error = std_dev / sqrt (n) ;
ds_init_empty (&vstr);
{
double T;
double n, mean, variance;
+ double std_dev;
+ double std_error;
- moments1_calculate (ws->totals, &n, &mean, &variance, NULL, NULL);
+ moments1_calculate (ws->dd_total[v]->mom, &n, &mean, &variance, NULL, NULL);
- double std_dev = sqrt (variance);
- double std_error = std_dev / sqrt (n) ;
+ std_dev = sqrt (variance);
+ std_error = std_dev / sqrt (n) ;
tab_text (t, 1, row + count,
TAB_LEFT | TAT_TITLE, _("Total"));
tab_double (t, 5, row + count, 0, std_error, NULL);
/* Now the confidence interval */
-
T = gsl_cdf_tdist_Qinv (q, n - 1);
tab_double (t, 6, row + count, 0,
mean + T * std_error, NULL);
/* Min and Max */
-
- tab_double (t, 8, row + count, 0, ws->minimum, fmt);
- tab_double (t, 9, row + count, 0, ws->maximum, fmt);
+ tab_double (t, 8, row + count, 0, ws->dd_total[v]->minimum, fmt);
+ tab_double (t, 9, row + count, 0, ws->dd_total[v]->maximum, fmt);
}
row += categoricals_total (cats) + 1;
for (v = 0; v < cmd->n_vars; ++v)
{
+ double n;
+
+
const struct per_var_ws *pvw = &ws->vws[v];
- const struct categoricals *cats = covariance_get_categoricals (pvw->cov);
const struct variable *var = cmd->vars[v];
const struct group_proc *gp = group_proc_get (cmd->vars[v]);
const char *s = var_to_string (var);
-
- const double df1 = pvw->n_groups - 1;
- const double df2 = pvw->cc - pvw->n_groups;
+ double df1, df2;
double F = gp->levene;
- tab_text (t, 0, v + 1, TAB_LEFT | TAT_TITLE, s);
+ moments1_calculate (ws->dd_total[v]->mom, &n, NULL, NULL, NULL, NULL);
+
+ df1 = pvw->n_groups - 1;
+ df2 = n - pvw->n_groups;
+ tab_text (t, 0, v + 1, TAB_LEFT | TAT_TITLE, s);
tab_double (t, 1, v + 1, TAB_RIGHT, F, NULL);
tab_fixed (t, 2, v + 1, TAB_RIGHT, df1, 8, 0);
for (v = 0; v < cmd->n_vars; ++v)
{
+ const struct per_var_ws *pvw = &ws->vws[v];
+ const struct categoricals *cats = covariance_get_categoricals (pvw->cov);
struct ll *cli;
int i = 0;
int lines_per_variable = 2 * n_contrasts;
++i, cli = ll_next (cli))
{
struct contrasts_node *cn = ll_data (cli, struct contrasts_node, ll);
- struct ll *coeffi = ll_head (&cn->coefficient_list);
- int ci;
+ struct ll *coeffi ;
+ int ci = 0;
double contrast_value = 0.0;
double coef_msq = 0.0;
- struct group_proc *grp_data = group_proc_get (cmd->vars[v]);
- struct hsh_table *group_hash = grp_data->group_hash;
-
- void *const *group_stat_array;
double T;
double std_error_contrast;
double df_denominator = 0.0;
double df_numerator = 0.0;
+
+ double grand_n;
+ moments1_calculate (ws->dd_total[v]->mom, &grand_n, NULL, NULL, NULL, NULL);
+ df = grand_n - pvw->n_groups;
+
if ( i == 0 )
{
tab_text (t, 1, (v * lines_per_variable) + i + 1,
if (cn->bad_count)
continue;
- group_stat_array = hsh_sort (group_hash);
-
- for (ci = 0;
- coeffi != ll_null (&cn->coefficient_list) &&
- ci < hsh_count (group_hash);
+ for (coeffi = ll_head (&cn->coefficient_list);
+ coeffi != ll_null (&cn->coefficient_list);
++ci, coeffi = ll_next (coeffi))
{
+ double n, mean, variance;
+ const struct descriptive_data *dd = categoricals_get_user_data_by_subscript (cats, ci);
struct coeff_node *cn = ll_data (coeffi, struct coeff_node, ll);
const double coef = cn->coeff;
- struct group_statistics *gs = group_stat_array[ci];
+ double winv ;
+
+ moments1_calculate (dd->mom, &n, &mean, &variance, NULL, NULL);
- const double winv = pow2 (gs->std_dev) / gs->n;
+ winv = variance / n;
- contrast_value += coef * gs->mean;
+ contrast_value += coef * mean;
- coef_msq += (coef * coef) / gs->n;
+ coef_msq += (pow2 (coef)) / n;
- sec_vneq += (coef * coef) * pow2 (gs->std_dev) /gs->n;
+ sec_vneq += (pow2 (coef)) * variance / n;
- df_numerator += (coef * coef) * winv;
- df_denominator += pow2((coef * coef) * winv) / (gs->n - 1);
+ df_numerator += (pow2 (coef)) * winv;
+ df_denominator += pow2((pow2 (coef)) * winv) / (n - 1);
}
sec_vneq = sqrt (sec_vneq);
n_contrasts,
TAB_RIGHT, contrast_value, NULL);
- std_error_contrast = sqrt (grp_data->mse * coef_msq);
+ std_error_contrast = sqrt (pvw->mse * coef_msq);
/* Std. Error */
tab_double (t, 4, (v * lines_per_variable) + i + 1,
TAB_RIGHT, T,
NULL);
- df = grp_data->ugs.n - grp_data->n_groups;
/* Degrees of Freedom */
tab_fixed (t, 6, (v * lines_per_variable) + i + 1,
projects / pspp / blobdiff